基于改进YOLOv5的电厂人员绝缘手套佩戴检测

电厂工作人员工作时需佩戴绝缘手套进行故障检修等任务,若未佩戴绝缘手套进行操作,将发生严重的电击事故。针对电厂内工作人员绝缘手套佩戴检测精度不高的问题,提出一种基于改进YOLOv5s(You Only Look Once v5s)的电厂内人员绝缘手套佩戴检测方法。该检测算法首先引入自校准卷积,有效扩大感受野,加强网络对弱特征的提取能力;其次加入注意力机制SK,让网络更加关注待检测目标;最后,将原YOLOv5s的损失函数替换为EIOU,来进一步提高网络对绝缘手套的检测精度。实验结果表明,相较于原始的YOLOv5s网络,改进后的网络提高了对绝缘手套佩戴的检测精度,其平均精度均值(mAP,mean Average Precision)提高了2.4%,证明了算法的实用性和高效性。

Detection of insulation gloves worn by power plant personnel based on improved YOLOv5

Power plant personnel need to wear insulation gloves during troubleshooting. If they do not wear insulation gloves, serious electric shocks may occur. Aiming at the problem of low detection accuracy of insulation gloves worn by workers in power plant, a detection method of insulation gloves worn by workers in power plant based on improved YOLOv5s (You Only Look Once v5s) was proposed. Firstly, self-calibrating convolution is introduced in the detection algorithm to effectively enlarge the receptive field and strengthen the ability of the network to extract weak features. Secondly, the attention mechanism SK is added to make the network pay more attention to the target to be detected. Finally, the loss function of the original YOLOv5s is replaced by EIOU to further improve the detection accuracy of insulation gloves. The experimental results show that compared with the original YOLOv5s network, the improved network improves the detection accuracy of insulation gloves wearing, and the mAP (mean Average Precision) is increased by 2.4%, which proves the practicability and efficiency of the algorithm